NEWS

---

ardldml 0.1.0 (2026-09-15)

• First release.
• dml_bounds() fits the DML-Bounds test of Villena (2026): a balanced, cross-fitted first stage orthogonalises the lagged levels against a high-dimensional control set, and the null of no level relationship is tested on the residualised variables.
• dml_bootstrap() implements the restricted system wild bootstrap, which regenerates the outcome and the focal regressor jointly under a shared Rademacher weight so the endogeneity channel is preserved.
• trend_absorption() and penalty_sensitivity() report whether a verdict survives a change of conditioning set or of penalty.
• classical_bounds_test() runs all three steps of the classical procedure and reads the statistic against a bracket simulated at the model's own sample size, rather than against a table calibrated at T = 1000.
• simulate_pss_bounds() regenerates the classical bracket from the data-generating process of Pesaran, Shin and Smith (2001); no critical-value table is stored in the package.
• Bundled passthrough data: nine monthly United States macroeconomic series, 1973-2020, so every example runs offline.